arangodb/Documentation/Books/Manual/Troubleshooting/Cluster

Cluster troubleshooting

• Cluster frontend is unresponsive

  • Check if the Coordinator/s in question are started up.
  • Check if the Agency is up and a leader has been elected. If not ensure that all or a majority of Agents are up and running.
  • Check if all processes have been started up with the same JWT_SECRET. If not ensure that the JWT_SECRET used across the cluster nodes is ideantical for every process.
  • Check if all cluster nodes have been started with SSL either dis- or enabled. If not decide what mode of operation you would like to run your cluster in, and consistently stick with for all Agents, Coordinators and DBServers.
  • Check if network communication between the cluster nodes is such that all processes can directly access their peers. Do not operate proxies between the cluster nodes.

• Cluster front end announces errors on any number of nodes

  • This is an indication that the Agency is running but either Coordinators or DBServers are disconnected or shut down. Establish network connection to or start the according nodes.
  • Make sure that the nodes in question share the same JWT_SECRET and SSL operation mode with the functioning nodes.